An honest account of PDF compression
Most of a large PDF is embedded images. Everything else — text, fonts, vectors, the document structure — is usually a small fraction of the total. That single fact determines what any compressor can achieve.
Lossless mode rewrites the file with object and cross-reference streams and clears the title, author, subject, keywords, creator and producer fields. Nothing is degraded. How much it reclaims depends entirely on how the file was written: on a well-produced PDF it will reclaim almost nothing, and on a very small file it can even add a few bytes. We show you the real number either way.
Strong mode re-renders each page as a JPEG. That is how this tool makes a PDF dramatically smaller in the browser, and it is genuinely lossy: the text layer is gone afterwards.
Which to use
If your PDF is a scan — photographed or scanned pages, no selectable text — strong mode is the right tool and costs you nothing you had. If your PDF came from a word processor and you can select its text, use lossless, accept the modest saving, and consider whether the file needs to be smaller at all.
Where a genuinely large reduction matters on a text document, the better route is usually to reduce what is in it: split out the pages you actually need to send. For images outside a PDF, the image compressor has far more headroom.